[
https://issues.apache.org/jira/browse/HADOOP-4258?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14731482#comment-14731482
]
Allen Wittenauer commented on HADOOP-4258:
------------------------------------------
* This likely won't be getting committed to trunk given that all development on
test-patch there has stopped. So please change this to use the HADOOP-12111
branch.
* preapply is too late. This should be in postcheckout. Doing that also
simplifies the code since returning 1 for failure will trigger all the cleanup,
etc., so that wouldn't have to be done in the plugin.
* I suspect the problem that [~owen.omalley] is actually talking about is a
problem where a new file was added because it conflicted with a pre-existing
file in the source tree. This patch would miss that situation.
* This seems like it would be very expensive on large patches due to the number
of calls to tr. Off the top of my head, it looks like it could be optimized to
use uniq -id first to limit the set of files that need to looked be at for at
least one of the loops (outer would likely be better, given point 2 above
should be handled prior to even running the inner loop!)
* Totally not a fan of having A and a and B and b as var names. I get the why's
but it makes it really hard to read at a glance! :D
Thanks!
> the test patch script should check for filenames that differ only in case
> -------------------------------------------------------------------------
>
> Key: HADOOP-4258
> URL: https://issues.apache.org/jira/browse/HADOOP-4258
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: build, test
> Reporter: Owen O'Malley
> Assignee: Jagadesh Kiran N
> Labels: newbie, test-patch
> Attachments: HADOOP-4258.001.patch
>
>
> It would be nice if the test patch script warned about filenames that differ
> only in case. We recently had a patch committed that had a pair of colliding
> filenames and subversion broke badly on my Mac.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
